BAe is the principal aerospace contractor in the U.K. and has divisions working with civil & military aircraft and guided weapons, amongst others. Their civil aircraft activity currently involves designing and building wings for the European Airbus projects and it is in the design of such aircraft that they wish to concentrate the efforts of the research.
SRC acts as the corporate research centre for the whole of BAe and contains six research departments employing approximately 200 people. Relevant to this project are the Aerodynamics & Vulnerability, Computational Engineering and Advanced Information Processing Departments. It is envisaged that, where possible within the scope of the research, the methods and experience that has been developed by these departments in the aerodynamic analysis of aircraft structures, the optimization of these structures and the construction of multi-disciplinary design methods and IT systems, be made available to the project. This support will be enhanced by the active collaboration of practising design staff from Airbus and MAA, who will supply the basic design models to be worked on and will help assess results as the project progresses. Staff from these divisions will also contribute at progress meetings and during the interfacing and testing of the codes being developed.
Rolls-Royce is a world leading power systems business, providing cost- effectively engineered products and services to commercial and military customers in propulsion, electrical power and materials handling markets around the world. Customers from the worlds leading airlines to executive jet operators rely on the powerful range of RR commercial aero engines and global support network. Military customers benefit from engines for helicopters, fast jets, trainers and transport aircraft, as well as naval vessels.
The research described here will take place in close collaboration with the Turbines section of the Engine Systems group within RR. This section has the responsibility for designing and manufacturing all turbine components - vanes, blades, discs and shroud liners. The project design work is centralised in its main offices in Derby and Bristol, dealing with predominantly Civil and Military engines, respectively. Generic research and methods for all turbine disciplines, for both civil and military engines, is coordinated across both these sites.
It is now common practice for large companies such as BAe and RR to outsource their IT activities and it is therefore helpful that RR's IT provider, EDS Ltd., are willing to collaborate in this programme of work. They are currently developing an engineering development network within RR which will be used to host some of the work described here. There staff will be actively involved in the work, both on a day-to-day basis and also at progress meetings.
The intelligent CAD software used in this work will be supplied by Knowledge Technologies International, inc., and their technical staff will also collaborate in the project by attending and contributing to the regular progress meetings scheduled into the project as will the staff of Applied Computing and Engineering Ltd., who will be the suppliers of some of the CFD codes used.
These developments in the design process have also given rise to a number of knock-on effects that are still not fully understood and which may have adverse consequences on the long term utility of designs produced in this way. In particular, it has been the normal practice of aerospace companies to design their products in families, whereby initial designs are produced with inherent `stretch' capability, so that second, third and further generation increments can be produced without re-engineering the whole product. i.e., as a form of adaptive design[21]. Such an adaptive approach often implies that judgements must be made concerning likely technological developments during the life of a design family. For example, in gas turbine design, inlet temperatures have risen some 400K over the last 40 years while typical engine families have had lives of at least 25 years[22]. Clearly, design decisions taken in the 1970's will have significant effects on engines running with 1990's technologies. Moreover, such technology changes will not be something that optimization based design tools can allow for without explicit guidance being provided: the previous practice of senior engineers has been to meet this need by considering key decisions alongside suitable design margins, etc., based on experience of past practice and perceived likely developments, in terms of both technical performance and market needs. There is a danger that this process will be jeopardized as increasingly sophisticated, goal-oriented design methods such as optimization are introduced.
In short, industry is currently using knowledge based and parametric CAD systems. It is also currently using evolutionary based search methods. There is now a need to produce designs by combining these techniques while maintaining stretch capability within design families. The use of these methods is thus a pressing problem now facing the aerospace industry and one where a timely programme of design research will be both novel and of significant value. The programme described here addresses this need, while also seeking to provide mechanisms whereby the results of design searches can be made more robust and also used to automatically adapt engineering knowledge bases.
The key element in getting this right is the assessment of which design assumptions, operational modes, material properties, technological limits, manufacturing processes, etc., (and even which components) can be taken as essentially unchanging from variant to variant and which cannot, and then to allow for this distinction in the design process[23]. Thus, in the 1960's and 1970's, when designing car engines it was common practice to engineer engine blocks such that they could be bored at different diameters even though this meant that the smaller capacity engines had over-sized blocks, simply because the capital costs of re- engineering each block as the capacity changed, say from 1800cc to 2000cc to 2500cc was prohibitive (this is, of course, an area where technological developments have subsequently revolutionised the design and manufacturing processes).
In the current era the main focus of reuse is on the design process itself rather than components and so when designing a civil transport wing, the design team have, for example, knowledge bases concerning the kind of trailing edge flaps that can be used to alter geometry during landing and take-off. Similarly, gas turbine designers have information on classes of blade cooling systems. When producing later variants such information will be reused. In addition, however, the design team will normally also assume that the achievable performance targets have drifted upwards by a few % over a period of years. They thus make decisions concerning the performance targets of next generation wings or blades based on these earlier trends. If such improvements are not assumed, conservative designs arise; conversely, if overly ambitious targets are set they may prove infeasible in practise. Moreover, design teams are increasingly looking to capitalise on latest generation adaptive IT methods such as knowledge bases and evolutionary search methods.
- the use of intelligent CAD systems containing design rules for a product family;
- multi-level search methods which can call on a variety of analysis codes;
- robust and adaptive information processing techniques and
- a multi platform meta-computing environment (using CORBA and STEP standards).
It must be stressed that the aim of the research is not to provide linkage between the CAD and search tools based on a simple "forward path" model: rather, the knowledge housed in the CAD system will be used to build search environments for the optimizers while the results of searches will be used to refine and add to the rules in the knowledge base, i.e., a "closed loop" approach.
To provide a realistic environment for carrying out such research, the analysis codes used must span the full gamut of available methods running from empirical data sets to 3D Euler methods for fluid mechanics and basic beam models to 3D FE codes for structural mechanics. The multi-level search aspects of the work will draw on the results of an existing EPSRC project (GR/L04733) and will be based on latest generation evolutionary methods, while knowledge based CAD methodology, drawing on the ICAD experiences of the BAe Airbus and RR Turbine sections, will be used to embed typical design family knowledge. Advanced computational loading strategies will also be vital in getting the best from the available computing platforms and this aspect of the work will draw on recent studies carried out under two EU funded design projects (Brite-Euram MDO project BE95- 2056 and Esprit project 20189 PROMENVIR).
In summary, the main objective of the work will be to gain insight into how knowledge based and robust design processes need to be used with stochastic search methods and the layers of analysis tools already available: in short we have plenty of sharp blades - we now need to make some scissors! This will be accomplished by implementing interfaces between an intelligent commercial CAD system (here ICAD), evolutionary search methods[32] and a realistic range of types and levels of engineering analysis software running on a multi platform meta-computing system. This process will be based on the CORBA and STEP approachs and CFD/FE codes of the type used by BAe and RR in the fields of structural and fluid-dynamic design. The resulting system will then be used to explore and evaluate modifications to design encodings, knowledge management strategies, robust design methods and multi-level search engines to enable multi-level and parallel interactions between the design methods. Almost all of the analysis tools used will be commercially available, general purpose codes so as to enhance the generic nature of the research.
The results of this work will be fed directly back into BAe and RR by the active collaboration of their staff during the project. It wil be based on commercial data-sets and codes and that the new methods developed be assessed by company design staff. This will be carried out by the implementation of suitable versions of the software on machines compatible with those used in the companies so that the methods developed can be made available in design offices and used by designers alongside their normal facilities. Such exercises will be backed up by secondments, seminars and extended training sessions carried out by the research staff of the project for the benefit of practising designers.
It is of course important that other industrial sectors can capitalise on the kind of generic design research described here. The inclusion of EDS in the project, who are an IT support provider that manages engineering IT facilities across a broad range of U.K. industries, will clearly help lead to greater understanding and exploitation of the techniques developed. This will span a variety of industrial sectors as EDS seeks to use best practice IT systems for all its customers. This process will be further aided by the participation of two software houses in the study since their individual products will be able to benefit from the work carried out and then made available to other user communities.
In addition, the research staff involved attend and give presentations at the industrial/academic seminars organized by the other research centres funded by the companies as well as the EDC's at Cambridge and Plymouth and also at relevant international conferences in this area. It also the normal practice of those involved to publish journal papers describing their research at regular intervals and to make their work available on the world wide web. Additionally, Prof. Keane and company staff are members of EvoNet - the European network for Evolutionary Computing, which will provide a further channel for technology transfer. Prof. Keane is also a founder member of EPSRC's engineering network in Adaptive Computing in Design and Manufacture - ACDMnet.
This work will be supervised by Prof. A.J. Keane acting as P.I. He will be assisted by two full-time post-doctoral research assistants, with skills in ICAD, FE/CFD, optimization and CORBA/STEP interfaces together with two research students working on the related but more speculative research topics. The first RA will mainly focus on the use of the CFD and FE tools in this context and carry out investigations into design representations, while the second RA will deal with the required ICAD code development and multi-platform CORBA/STEP compliance issues. Both will be involved with testing and interaction with BAe and RR staff. The students will tackle the advanced representation and neuro-fuzzy/Bayesian modelling issues and will, of course, be primarily focused on completing PhD theses. Their activity will provide a good deal of support to the main project and forms an important aspect of the project.
1. A. J. Keane, W. G. Price, and R. D. Schachter, ``Optimization Techniques in Ship Concept Design,'' Trans. R.I.N.A. 132(A)(1990).
2. A. J. Keane, ``Experiences with optimizers in structural design,'' pp. 14-27 in Proceedings of the Conference on Adaptive Computing in Engineering Design and Control 94, ed. I. C. Parmee,P.E.D.C., Plymouth (1994).
3. A. J. Keane, ``Passive vibration control via unusual geometries: the application of genetic algorithm optimization to structural design,'' J. Sound Vib. 185(3) pp. 441-453 (1995).
4. A. J. Keane, ``Genetic algorithm optimization of multi-peak problems: studies in convergence and robustness,'' Artificial Intelligence in Engineering 9(2) pp. 75-83 (1995).
5. C. M. Fonseca and P. J. Fleming, ``An overview of evolutionary algorithms in multiobjective optimization,'' Evolutionary Computing 3(1) pp. 1-16 (1995).
6. A. J. Chipperfield and P. J. Fleming, ``Genetic algorithms in control systems engineering,'' Control and Computers 23 pp. 88-94 (1996).
7. A. J. Chipperfield and P. J. Fleming, ``Multiobjective gas turbine engine controller design using genetic algorithms,'' IEEE Trans. on Industrial Electronics 43(5) pp. 583-589 (1996).
8. S. Dwarakanath and K. M. Wallace, ``Decision making in design: Observations from design experiments,'' J. Eng. Design 6(3) pp. 191-206 (1995).
9. K. M. Wallace, ``Product development and design research,'' in Proc. 11th Int. Conf. on Engineering Desifn (ICED 97), Keynote address, Tampere (1997).
10. T. N. S. Murdock and K. M. Wallace, ``An approach to configuration optimisation,'' J. Eng. Design 3 pp. 99-116 (1992).
11. C.J. Harris, Advanced Adaptive Control, J. Wiley (1995).
12. C.J. Harris, Neuro fuzzy adaptive modelling and control, Prentice Hall (1994).
13. C.J. Harris, Intelligent Control: Aspects of fuzzy logic and neural networks, World Scientific Press, (1993).
14. M. B. Giles and R. Haimes, ``Validation of a numerical method for unsteady flow calculations,'' J. of Turbomachinery 115(1)(1993).
15. P. Crumpton and M.B. Giles, ``Aircraft Computations Using Multigrid and an Unstructured Parallel Lirary,'' AIAA 95-0210(1995).
16. G. N. Shrinivas and M. B. Giles, ``OGV tailoring to alleviate Pylon- OGV-fan interaction,'' ASME 95-GT-198(1995).
17. C. Ruiz and R. Duffin, ``Experimental study of soft body impact at subordnance velocities,'' pp. 143-148 in Experimental Mechanics, ed. Allison,Balkema, Rotterdam (1998).
18. C. A. Featherston and C. Ruiz, ``Buckling of flat plates under bending and shear,'' Proc. I. Mech. Eng. 212(C) pp. 249-261 (1998).
19. C. Ruiz, S. R. Hallett, and S. Gungor, ``Application of fracture mechanics criteria to laminated CFRP beams under impact,'' pp. 118-124 in Proc. ECCM-8, , Naples (1998).
20. M. A. Hale, J. I. Craig, F. Mistree, and D. P. Schrage, ``DREAMS and IMAGE: A model and computer implementation for concurrent, life- cycle design of complex systems,'' Concurrent Eng. - Res. and Appls. 4(2) pp. 171-186 (1996).
21. P. C. Stadzisz and J. M. Henrioud, ``An integrated appraoch for the design of multi-product assembly systems,'' Computers in Industry 36 pp. 21-29 (1998).
22. A. G. Dodd, ``Mechanical design of gas turbine blading in cast superalloys,'' Materials Science and Technology 2 pp. 476-485 (1986).
23. S. Shahpar and L. Lapworth, ``A Forward and Inverse Three- Dimensional Linear Design System for Turbomachinery Applications,'' pp. 377-382 in Proceedings of the 4th European Computational Fluid Dynamics conference, J. Wiley, Athens (1998).
24. S. W. Sanderson, ``Cost models for evaluating virtual design strategies in multicycle product families,'' J. Eng. & Tech. Mang. 8 pp. 339-358 (1991).
25. K. Yamamoto and O. Inoue, ``Applications of genetic algorithms to aerodynamic shape optimization,'' pp. 43-51 in Proceedings of the 12th International Conference on Computational Fluid Dynamics, AIAA, San Diego (1995).
26. KTI inc., The ICAD System, http://www.ktiworld.com/products.shtml (1998).
27. M. Brown, S. R. Gunn, C. Y. Ng, and C.J. Harris, ``Neurofuzzy systems modelling: A transparent approach,'' in Dealing with Complexity: A Neural Network Approach, ed. K. Warwick,Springer Verlag, London (1997).
28. R. J. Buck and H. P. Wynn, ``Optimization Strategies in Robust Engineering Design and Computer-Aided Design,'' Quality and Reliability Engineering International 9 pp. 39-48 (1993).
29. A. J. Keane, ``A Brief Comparison of Some Evolutionary Optimization Methods,'' pp. 255-272 in Modern Heuristic Search Methods, ed. V. Rayward-Smith, I. Osman, C. Reeves and G. D. Smith,J. Wiley (ISBN 0 471 96280 5) (1996).
30. M. E. Topliss, C. A. Toomer, and D. P. Hills, ``Rapid design space approximation for two-dimensional transonic aerofoil design,'' in AIAA 34th Aerospace Sciences Meeting, , Reno (1996).
31. M. A. El-Beltagy and A. J. Keane, ``A Comparison of Genetic Algorithms with Various Optimization Methods for Multi-level Problems,'' Eng. Appl. of Artificial Intelligence, (to appear).
32. A. J. Keane, The OPTIONS Design Exploration System User Guide and Reference Manual, http://www.soton.ac.uk/ ~ajk/options.ps (1994).
This article may be found at http://www.soton.ac.uk/~ajk/utp_fps.html